JP2009054174A - Method for supporting product selection by consumer - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To present structured data related to attributes of a product, and to allow a user to perform optimal selection according to one's own personal profile. <P>SOLUTION: In this method, the data related to taste and requirement to the plurality of attributes of the product inside a product domain are received from the user, and a set of the products to the user is presented according to the data. In the method, a list of items is created according to the taste and the requirement of the user about the attribute A(n) of the item inside the product domain. The method includes steps: for defining a set of available items about the attribute A(n); for making the data showing the taste or the requirement about the attribute A(n) be input; for storing the input data; for determining a residual item set satisfying the requirement related to a subset of the attribute A(i) (i=n, n-1, etc., 1); and for allocating a score to the item inside the residual item set according to the input data by repeating the above steps. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a method for assisting a user in making a decision based on a user profile, and more particularly, a product domain such as a car purchase based on a user specific request or product selection preference. The present invention relates to a method using a computer system that assists in making a complicated selection from a set of products.

  Currently, considerable efforts are being made to develop computer applications in market transactions and consumer product distribution. For example, U.S. Pat. No. 4,775,935 to Yourick, U.S. Pat. No. 5,041,972 to Frost and U.S. Pat. No. 5 to Stack. , 124,911 describe a system that processes information about the product that the consumer selects to determine such product attributes that are important in the consumer's decision. Such a system helps manufacturers design products that are more likely to sell, and helps market traders understand what products are sold to which consumers. However, such products are not focused on helping consumers make complex product choices.

There is also considerable development in the field of decision-making systems. For example, US Department of Defense publication No. T998,008 by DeLano Jr., US Pat. No. 4,829,426 granted to Burt, and Alexander et al. U.S. Pat. No. 5,182,793 describes a software-based system for a computer that is a tool for a decision making process. However, such a system does not provide an easily available technique that allows the user profile to be applied to the decision making process in intelligently selecting the user profile.
US Pat. No. 4,775,935 (Yourick) US Pat. No. 4,829,426 (Burt) US Pat. No. 5,041,972 (Frost) US Pat. No. 5,124,911 (Stack) US Pat. No. 5,182,793 (Alexander et al.) US Department of Defense Publication T998,008 (DeLano Jr.)

  Thus, there are quite a few technologies that apply computer systems to decision making, but such technologies solve complex choices such as consumer product selection based on the unique profile of the person who chooses, and There is a need to apply in effectively providing results to that person. This need is particularly evident in the field of consumer purchase decisions coupled with “expensive item” purchases that have traditionally been offered through untrustworthy or prejudice sales channels. Purchasing such expensive items generally requires a fair amount of analysis and logic because a great deal of choice and many variables must be considered in decision making Because.

  The present invention processes information for identifying a product within the product domain for the user and presents the structured data regarding the attributes of the product within the product domain to the user in an easily understandable and efficient manner. This provides a system that allows the user to make the best choice according to their profile. Thus, the present invention includes a user interface that presents a sequence of input prompts prompting the user to collect preference and request data for attributes of multiple products within a product domain. The decision engine is coupled to a user interface that filters the product domain and presents a group of products to the user for product selection according to the collected preference and request data. The preference data includes variables associated with specific attributes specified by the user to have a degree of relevance to product selection within the product domain, but is not an absolute requirement. The request data includes variables associated with specific attributes specified by the user that are or are not required for product selection within the product domain. In accordance with the present invention, the user is guided by a sequence of input prompts to assist in creating a personal profile regarding preferences and requirements for products within a particular product domain. The decision engine applies the above requirements and preferences to generate a list of one or more products that meet the requirements specified by the user profile. This product list is ranked according to preferences specified by the user profile to assist the user in selecting products.

  The user interface according to one aspect of the present invention includes a navigation window, which indicates the position to the user with a sequence of input prompts, which in turn allows user feedback regarding the current position. The navigation window also responds to previous prompts in a sequence of prompts in one alternative, to show the user how much the product domain is narrowed at that prompt as additional feedback in the product decision-making process. Has been used.

  A decision engine according to one aspect of the invention stores a plurality of attributes in a hierarchical array. Resources in the decision engine assign weights to corresponding attributes in the plurality of attributes in response to the collected preference data and the hierarchical arrangement of the corresponding attributes. A set of attribute measurements is generated based on the collected data. This measurement is used to calculate a product score for each product in the product domain that satisfies the required data.

  According to another aspect of the invention, attribute measurements of attributes in a plurality of attributes include (1) a numeric type of numeric evaluation of the attribute, and (2) a Boolean type of attribute that can be characterized by a true / false indicator, (3) enumerated types for attributes selected from a set or list, (4) optional types of attributes included as reference features that are included or not included as optional features, and (5) numerical ranges. Having individual measurement types selected from a set of measurement types including a ranged numeric type for the specified attribute. According to this aspect of the invention, the decision engine resource calculates a product score in response to a normalization routine selected according to the measurement type of the corresponding attribute.

  The decision engine also includes a resource that assigns rules to attributes, which in the process of generating attribute measurements and product scores specify responses to missing attribute input data, and the input data is missing The rule is applied to the identified attribute at the time of. The decision engine also includes resources that assign rules to the attributes in response to irrelevant input data for the attributes and adapt such rules as appropriate when generating attribute measurements and pass / fail scores. The data may be missing, for example, due to an input error or the user skipping certain prompts during the input process. Such data is irrelevant to that attribute and that product, for example, if that particular attribute does not apply to a particular product.

  According to yet another aspect of the present invention, a sequence of input prompts includes a slider bar displayed on a display positioned by the user to indicate a selection within range on the input device, and a caret on the slider bar. It includes at least one input prompt that presents a selection presented by a graphic such as a selector element such as a symbol. The resource coupled with the decision engine computes a numerical value in response to the user's selection. Also, such resources in some embodiments use non-linear features of certain attributes of products in the product domain to calculate numerical values based on fuzzy logic functions.

According to one alternative, the invention features a user for a set of attributes A (n) organized in a sequence for an index (n) equal to 1 to N of the attributes in the product domain. There is a way to generate a list of items from the product domain according to your preferences and requirements. This method includes (although not necessarily in this order) the following steps:
(1) defining a set of available choices for attribute A (n), the set of available items for this attribute A (n) being in the product domain for the first attribute (n = 1) Including available choices for all items in the list and including available choices for items in a set of remaining items if n is greater than one;
(2) prompting the user to enter data indicating preferences and requests for attribute A (n) in the set of attributes according to the set of available choices for attribute A (n);
(3) storing input data for the attribute A (n) and developing a user profile;
(4) determining a set of remaining items that satisfy a request for a subset of attribute A (i) (where i is a set of remaining items that satisfy a request for a previous attribute in the set; (Equal to n, n−1,..., 1 to correspond to items in the product domain);
(5) providing the user with an indicator that indicates the position in the sequence and the count of items in the set of remaining items;
(6) for n equal to 1 to N in the set of attributes, defining, prompting, storing, determining, and providing for attribute A (n); And (7) assigning a score to the items in the set of remaining items according to the input data indicating the preference for the attribute in the attribute set, and ordering the items in the set of remaining items according to the score.

  Thus, according to the method of the present invention, the user is presented with a sequence of prompts and provides data regarding preferences and requests in response to prompts in the sequence. As each prompt is completed, a set of remaining items is calculated based on the specified request, the user is notified of the current position within the sequence of prompts, and a count of the set of remaining items is presented to the user. The In this way, the user navigates through a sequence of prompts, identifies preferences and requirements for specific attributes, and progressively narrows the selection of items within the product domain until a manageable set is provided. It becomes possible. This allows the user to make intelligent product selections based on a manageable set of items that satisfies the user's personal profile for the product being purchased.

  Thus, the present invention is designed to use a consumer's personal profile when determining items from a large number of multi-faceted items in a domain (eg, cars, mutual funds, news sources, universities, etc.). Provide an engine and process for optimized decisions. The present invention guides the consumer in the process of creating a personal profile by expressing a high level abstract statement and a low level explicit selection statement that the consumer wants. This profile is represented as a hierarchical collection of requirements and preferences, but the requirements here are strong constraints that must be satisfied in order to stay within the set under consideration for an item. , Given as a degree of interest that indicates a consumer's relative dislike or dislike for a particular attribute. The personal profile is then applied to the product domain using a normalization routine to generate a ranked list of items that satisfy the order ordered by the user expressed preferences. In this way, representatives representing a low level of the consumer's profile based on demand and preference and the domain of interest can be personalized to the consumer using abstract hierarchical and statistical representatives. An analytical decision processing tool is provided for determining a ranked item list. This system is easy to use and significantly improves the efficiency of the marketing channel for the product domain.

  Other aspects and advantages of the invention will be apparent from the following description of the drawings, detailed description and claims.

  A detailed description of the present invention is provided with reference to FIGS. 1-15, in which FIG. 1 illustrates the present invention in the context of an Internet server. Thus, the system of the present invention includes a decision engine server 10 such as a computer workstation having decision engine software loaded therein according to the functions described below. The decision engine server 10 includes a display 11 on which a graphical user interface is displayed for the purpose of collecting user profile data from persons local to the decision engine server 10. In addition, the decision engine server 10 is coupled to a wide area network such as the Internet 13 by the connection unit 12. User stations 14, 15, 16 are coupled to the Internet 13 by connections 17, 18, 19, respectively. Each user station 14, 15, 16 includes a display 20, 21, 22 on which a graphical user interface is presented for collecting user profile data under the control of the decision engine 10. In this way, a user at the user station 15 can access the decision engine 10 using standard Internet protocols. The decision engine server drives a graphical user interface on the display 21 of the station 15 to collect user profile data and present the user profile data to the decision server.

  In accordance with the present invention, the decision engine server 10 includes a data structure that defines a domain product collection 25, which includes a list of items in a product domain, such as cars available on the market, and this list. It consists of specific product information about the attributes available for each item in the set. Additional data structures are included with the server 10 that defines an attribute set 26 for the product domain. The domain attribute set 26 in the preferred system consists of a hierarchical array of attributes specified for the entire product domain 25.

  User profile information is collected from the user station and stored in the user profile data structure 27, which is a request and preference feature according to the input data provided by the user. The decision engine server 10 processes the user profile 27 according to the domain attribute set 26 and the domain product set 25 to generate a ranked list of the remaining items in the set for presentation to the user.

  The environment of FIG. 1 illustrates the use of a decision engine in setting up the Internet. Of course, the decision engine is applicable to local installations such as intranets and single workstation embodiments.

  FIG. 2 is a functional block diagram of the decision engine and consumer profiling process according to this invention. This system includes a decision engine 100, a question and answer sequence module 101, a graphic user interface driver 102, and a database module 103. The decision engine module 100 includes an analysis decision processing module that uses a low-level display consumer profile and an abstract hierarchical and statistical display interest domain to determine a personal ranked item list for the consumer. It is out. The question and answer sequencer 101 uses editorially described narratives to guide consumers dynamically and personally by creating a personal profile. Database module 103 is a relational database or other storage system used for storage and maintenance of consumer profiles. The graphical user interface driver 102 includes simple graphical primitives and other tools that allow the user to select and express opinions in response to input prompts from the question and answer sequencer 101.

  The question and answer sequencer 101 in the preferred embodiment includes a first page 105 for the attributes of the first set, a second page 106 for the set A, an attribute set B, and a third page 107 for the attribute set C throughout the attribute hierarchy. HTML (Hypertext Markup Language) scripts that specify the order of the set of pages 104 are included. A page 108 is also included to present the results to the user. The question and answer sequencer 101 provides an HTML page according to the script to the graphic user interface driver 102 that provides an image and prompts the user to provide input data. In addition, the question and answer sequencer 101 tracks the current processing position in the script. Thus, the graphical user interface driver includes a navigation window 110 that indicates to the user the position in the sequence of prompts. The graphical user interface driver 102 is coupled on line 111 to a display and input device to collect preference and request input data. As described above with reference to FIG. 1, line 111 is a direct connection to a local display, as appropriate to the needs of a channel or specific implementation to reach a remote user via the Internet or other network link. It may contain things.

  The decision engine 100 is represented by an engine parameter structure 115, an attribute hierarchy 116, an item set 117, a measured attribute vector 118, and a residual list of items and scores 119 (eg, fields in the item set structure 117). Contains a set of data structures. Using the data structures 115-119 and profile data output on the line 120 from the question and answer sequencer 101, the decision engine calculates weighted attribute values in the set, and missing data for the attributes in the set. Process unrelated data, determine the residual item set based on the request previously represented by the user in the question-and-answer sequence, determine the scope of the attribute based on this residual set, and determine the residual based on the preference value Rank a set of items.

  The engine parameter structure 115 includes information such as normalization routines for specific measurement types and other global parameters of the decision engine for specific domains. The attribute hierarchy 116 consists of a set of nodes arranged in a hierarchy for attributes that are characterized in the user profile. The item set 117 consists of a list of all items in the product domain and characteristics related to the specific attributes supported for each item. The measured attribute vector 118 includes a normalized value of the consumer profile for the attribute hierarchy based on input information provided by the user. The residual list and score 119 provides a result based on the measured attribute vector that remains applied to the item set. Once the prompt processing is complete, this data structure 119 is used to provide the user with a resulting set of items from the product domain that matches the user's preferences and requirements.

  FIG. 3 is an architecture diagram of the decision engine 100. The decision engine includes an attribute hierarchy 116, an item set vector 117, and a measured attribute vector 118. A measured attribute / item matrix 150 is also involved in the decision engine processing.

  The attribute hierarchy 116 is specified by a data structure of nodes arranged in the hierarchy array. Thus, this hierarchy 116 is a hierarchical organization of attributes related to domains that can be considered by consumers. This hierarchical organization or other organization of attributes is based on knowledge engineering within the domain and expert opinions within the domain. In this way, this organization is determined heuristically and optimized for a specific product domain. In the preferred system, this hierarchy is represented as a tree structure that includes a root 151 that represents the entire domain (eg, a car). Internal nodes down the tree represent progressively more specific composite attributes. For example, the car domain that is one level below the route 151 has a composite attribute node 152 for the size, and an attribute node 153 for the external size is one level below the node 152. Leaves 154, 155, and 156 represent specific measurable attributes for the item. For example, in a car, the leaves 154 to 156 for the external size include height, length and width. Another internal node that is one level below the car size node 152 is a node 157, which corresponds to the internal margin. The leaves of the node 157 include a leaf 158 for the head margin and a leaf 159 for the leg margin.

  A schematic diagram of the attribute hierarchy 116 is shown in FIG. 3, but this hierarchy can be very complex because it covers many attributes of a complex product.

  According to the present invention, the attribute hierarchy also includes a priori weighting for components of the composite internal attribute. For example, the a priori weighting for the external size is 33% high, 33% wide, and 33% long. Both the abstract organization of attributes in the tree and the corresponding a priori weightings are based on knowledge engineering in the domain, the opinions of domain experts, and statistical analysis. An attribute hierarchy is an abstract representation of the organization of features considered within a domain and a generic representation of the relative importance of each feature.

  The measured attribute vector 118 is used to record the statistical characteristics of each measured attribute for the complete population of items. That is, each cell in vector 118 represents a complete description of one attribute across all items. This description includes parameters such as the minimum value, maximum value, average value, intermediate value, standard deviation value, and linear approximation value of the measured attribute distribution. In addition, the cell contains instructions regarding how user input should be interpreted at run time.

  A measured attribute vector is associated with a leaf node in the attribute hierarchy and is used along with the user's profiling information to generate attribute measurements that are recorded to define the user profile.

  There are five types of measured attributes supported by the decision engine: (1) numeric values, (2) Boolean values, (3) enumerations, (4) opinions, and (5) ranged numeric values. The numeric measured attribute is a floating point evaluation of the feature, for example the car domain numeric attribute contains the maximum speed and brake distance. A Boolean measured attribute is an attribute that can be characterized by authenticity, for example, whether a car is important or not. An enumerated measured attribute is a collection or list representation of attributes, such as a list of colors available for a car. The measured attribute vector has each of the enumerated elements obtained from the list in the vector, for example each of red, green or blue is included in this vector. At construction time, the enumerated elements are adjacent to each other in the vector, the first element is marked as the first in the enumeration list, and the last element is marked as the last element in the group. There are two types of enumerations: a mutually exclusive enumeration and a non-mutually exclusive enumeration. In the case of mutual exclusion, only one of the listed elements can be true for an item. Opinion measurement types are similar to Boolean types, except that they have one false state and two true states. The false state implies that the feature is not available. The first true state indicates that the feature is available as a standard device. The second true state indicates that the feature is available as an add-on option for the item. The ranged numeric attribute is used for a numeric attribute that cannot be expressed by one numeric point and covers a numeric range.

  The measured attribute vector 118 is used to form a measured attribute / item matrix in response to the user profile so that the user profile is used to form the attribute hierarchy. The measured attribute / item matrix 150 includes a normalized value for each measured attribute for each item in the domain. Each entry is a normalized floating point value in the range of 0.0 to 1.0 in the preferred example. The method used for normalization depends on the measured attribute type as shown in Table 1 below. If measured attribute data is not available for a particular item, it is marked as being in the corresponding matrix. There are two types of data that are not available: missing data and unrelated data. Missing data is information about items that should be available but are not currently available for some reason. Unrelated data is information that is not related to a particular item (eg, a four-wheel drive transmission for an item that is only available in the case of a two-wheel drive transmission).

  The decision engine processes missing data according to the rules specified for each attribute. Thus, for example, when data for a specific attribute is missing, it is possible to price or ignore that attribute with a neutral value when assigning a score for an item. Alternatively, missing data may be characterized as a request by a particular attribute associated with it.

  The item vector 117 is simply a list of items in the product domain, specifications that indicate whether they are included or excluded in the remaining item list during processing, and a floating point score accumulated during engine execution. , Is made up of. The set of remaining items and scores is thus determined by the fields in the item vector 117.

  During runtime, the engine uses the attribute hierarchy 116, the measured attribute vector 118, and the item vector 117 together with the user profile that forms the attribute hierarchy 116 to determine personalized results. The engine has the ability to run in both the item direction, where the user's attribute profile determines the ranked list of items, and the attribute direction where the item list generates the scope of attributes used during the prompting process. ing.

  While traveling in the item direction, the engine receives both user attribute profiles and engine operating instructions, including propagation methods used and missing data and other anomaly handling methods. The first step in running the engine in the item direction involves creating a measured attribute vector according to the user profile using a personalized instance of the attribute hierarchy. This operation involves an operation in which the preference value descends in the attribute hierarchy tree and normalizes the request at the leaf of the attribute hierarchy. Although preferences may be stated anywhere in the hierarchical attribute tree, requests are specific measured attributes and are therefore only recorded at the leaves of the attribute hierarchical tree. This step is described by a non-leaf node in the user profile or descends the leaf and propagates using a propagation algorithm.

  According to the first embodiment, the operation of propagating preferences to the leaf nodes of the attribute hierarchy is performed by a second alternative using an optimal algorithm and further according to a braked algorithm. An exemplary optimal braked algorithm is shown in Table 2 below.

  The propagation algorithm is specified as part of the operational instruction that passes through the engine. Optimal algorithms allow the user to treat certain detailed attributes as the equivalent of a larger and wider attribute concept. The damped algorithm adjusts preferences according to individual positions in the attribute hierarchy, thereby preventing any detailed attribute preferences from overwhelming wider high level attribute preferences.

  Also in this step, the request stated in the leaf attribute in the profile is converted to its normalized value. The normalization method for each attribute in an embodiment is the same as the method used to create the measured attribute / item matrix.

  Next, the value of the field in the item vector, indicating inclusion or exclusion and score, is generated using the measured attribute vector and the measured attribute / vector matrix instance. Each cell in the item vector represents one item in the domain. If the item satisfies all requirements, the corresponding item cell is marked as “included”. In addition, for each possible item, the corresponding cell contains a score in the normalized embodiment ranging from 0.0 to 1.0. The engine also checks the request at the same time and calculates a preference ranking for each item as a whole. The engine progressively moves down the measured attribute vector instance to check that each item in the matrix meets local requirements. If the item does not meet the requirement, it is marked as “excluded” and removed from future calculations. If the data element is found missing or irrelevant, the engine uses a data anomaly processing instruction. These data anomaly processing instructions specify the number of missing and / or irrelevant data elements or the type of missing and / or irrelevant data, which can be given with the expectation that the calculation will be successful. . As the engine travels, it tracks the type and number of missing and unrelated data elements encountered for each item. If these numbers exceed a specified value, or if a type is encountered in an abnormal handling instruction, the item is marked impossible.

  In addition to request processing, the engine calculates item rankings based on user preferences. If the item is possible, the margin contribution for this attribute, ie its normalized measured attribute value multiplied by the calculated preference for the attribute in the measured attribute vector, is added to the item score in the item vector. . Finally, the engine normalizes the item score and classifies the item vector. Item scores are normalized, for example, so that the best scoring item receives a score of 1.0 and the scores of all other items are divided by this best score to map each to a 0 to 1 interval. . Furthermore, the set of possible items in the vector is sorted in decreasing order such that the best scoring item appears in the first cell. Impossible items (marked “excluded”) will appear at the end of the picture section. Results are thus created and presented to the user to indicate items that satisfy the requirements ranked according to the user's preferences.

  The decision engine also runs in the attribute direction during the prompt process as described below. In the attribute direction, the effective range of the attribute is generated using a list of items. Here, the engine goes through a list of item vectors with the item cells of interest marked for inclusion and the attributes to be considered. The engine travels down the input vector and determines the effective range of each attribute of interest. These scopes are then returned to a call function such as a prompt process as described below.

  FIGS. 4-14 illustrate an implementation of a preferred embodiment of a graphical user interface that presents a sequence of prompts to collect preference and request data to create a personal profile. FIG. 4 shows a screen breakdown according to a preferred embodiment. This screen has a title in area A, which is generally a single word indicating a topic on the screen. This screen also has an explanation field in area B. This description field provides a brief description of potential tradeoffs on the current topic. The text generally provides a sufficiently detailed description so that the user does not automatically select the most preferred settings for all categories.

  This screen also includes a user input widget in field C. The field C widget is a tool for providing input by the user, and takes various forms as shown in FIGS. In the example shown in FIG. 4, two sliders that present the positive preference feature shown in FIG. 13 are shown. This screen also includes an “exactly specified” field in its area D. This field is a checkbox that is available whenever the user can provide a specific request at a more detailed level than the topic on the current screen. By setting the correct designation button, a user interface such as a screen as shown in FIG. 14 is shown to the user, and more detailed selection is possible.

  The screen of FIG. 4 also includes a navigation window, ie a bar in region E. The navigation bar in area E contains a list of screens specified by the title in area A on the screen. Thus, in the illustrated example, the screen used in the prompt sequence includes a first tab 200 for all cars, a second tab 201 for car features, a third tab 202 for car safety attributes, Are represented by tabs including a fourth tab 203 for the first dimension, a fifth tab 204 for the car performance and a sixth tab 205 for the technical attributes of the car. Each tab corresponds to an attribute set presented by a prompt screen such as the performance screen of FIG.

  After the user provides input in accordance with a preferred embodiment of the present invention, a count value of the number of remaining items in the domain is presented adjacent to each tab in the navigation bar E. Thus, for the first tab 200, the number of remaining items is shown as 728, representing all cars in the domain. In this example, in the second tab 201, the user has selected to narrow the field to about 600. In the third tab 202, the user has selected to narrow the field to about 400. In the fourth tab 203, the user has chosen to narrow the field to about 350. The user specifies a request and preference by making a selection on the fifth tab 204 for performance that further narrows the selection. In this way, the system always gives feedback and progress on the user's position with a sequence of prompts specified according to the script on the page, narrowing the choices made according to the requirements specified on the previous prompt page.

  Figures 5 through 13 show representative widgets used by the present invention for purposes of specifying requirements and preferences. Thus, FIG. 5 shows a check box. This check box is used when the user can select one or more options from the list. This box is checked by clicking on the box with the user input tool to select or not select an item. In general, checkboxes are used to specify requests in attribute sets.

  FIG. 6 shows a numerical slider. The numeric slider includes a bar and a caret that the user can position along the bar. The minimum value is displayed at the left end 210 of the bar. The maximum value is displayed at the right end 211 of the bar. The value of the area 213 is displayed differently depending on the position of the caret 212 along the bar. With the numeric slider, the user can set the minimum or maximum value of the attribute. In this example, the user has specified a maximum price. In an alternative embodiment, two carets are used on the same slider bar to allow the user to set both maximum and minimum values. The value that the slider increments in response to the movement of the caret is set by the designer according to a specific implementation. In addition, during the caret 212 feature, it slides evenly across the bar and stops at a specific location in the bar that corresponds to the actual selection available for the enumerated attributes. It is possible to set.

  FIG. 7 shows radio buttons. Specify attribute selections that are mutually exclusive using radio buttons. The radio button is set by the user clicking the button with the pointer tool.

  FIG. 8 shows a combo box widget. A combo box widget is similar to a pull-down window that allows the user to click the down arrow in a rectangular box to see an option list. In general, an empty string is available to the user, which allows the user to refuse to answer.

  The widget shown in FIGS. 9 to 11 allows the user to express both requests and preferences. In this way, the request can prompt for input at both ends of the widget, only on the left side or only on the right end.

  FIG. 9 shows a fully notched slider widget. By selecting the leftmost button according to this widget, the item containing the feature is dropped from the list. With the button on the far right, the item must be included so that the feature remains on the list. Pushing the “don't care” button in the center will leave a neutral setting for the future. If you set dislike, a negative preference value is set for the future. This setting prevents items from being dropped from the list. Rather, the information lowers the rank of the item from the list. The “Like” setting sets a positive preference for the feature and increases the ranking of items in the list. The button of FIG. 9 is preferred on pages where the list of features is very long and the graphic slider uses too much memory in the system displaying it. In some systems, preference values assigned in this way using fuzzy logic are: 1) Brule's "Fuzzy System-Tutor, A" (1985) and 2) Elkan's August 1994. It is processed using techniques such as those described in the IEEE excerpt "The Paradoxical Success of Fuzzy Theory".

  FIG. 10 shows a “not wanted” half-notched slider. This type of slider works with mutually exclusive enumerations, that is, attributes that share their survival with other attributes. For example, you cannot have a car that is both a coupe and a sedan. Thus, this tool provides an elegant way for the user to drop items from the list or to set preferences for acceptable items for the user. Thus, in the widget of FIG. 10, the request is set at the left end of the slider, while the preference is set as the caret moves to the right.

  A “must-have” notched slider is shown in FIG. This slider operates on features that can share their survival with other features. For example, a mutual fund can have both an automated investment plan and a check draft. In this way, the combo slider does not have the option of not wanting, but at the right end is given the option of having to convert the corresponding attribute into a request.

  Figures 12 and 13 show a preference-only widget that only allows the user to rank the attributes in the list. If a widget is used for attributes as much as you like, no items are dropped. Thus, FIG. 12 shows a positive / negative preference slider. The neutral setting for this slider is the center of the bar. When the caret is slid to the left, the user shows a negative preference, but in this example, the correlation with the negative preference is small. Slide the slider to the right to express your positive preferences.

  In FIG. 13, a positive preference slider is shown. The neutral setting for this slider is at the left end. Slide the caret symbol to the right to express more positive preferences. Negative preferences cannot be set with this bar. It is clear that the bar can be reversed to set a negative preference.

  FIG. 14 shows a set-specific screen used in combination with the screen shown in FIG. 4 in response to the exact designation set in the area D of FIG.

  Thus, more specific attributes of the performance category are specified in FIG. A navigation bar in region E of FIG. 4 is also shown in region 250 of FIG. This bar is modified only if there is an indication that the attribute performance set is being processed on a set-specific screen. Otherwise, the navigation bar remains constant because more specific attributes are processed in the set. Thus, the screen 14 includes a title field 251, a description field 252, and a widget field 253 used to specify more detailed preferences and requests.

  In the alternative system, the count value of the remaining items is updated each time the request is expressed on the set-specific screen.

  FIG. 15 shows a simplified flowchart of a process for generating a personal profile and providing a list of items from a decision process. This process includes the step of presenting a start page to the user, prompting for input related to attribute set X (a set may include one or more members), and providing an updated navigation window Start at 300. The system then accepts user input and indicates a measure of the attribute set X (step 301). The item / attribute preference score and request for set X is then calculated (step 302).

  After examining the item / attribute preference score for attribute X, the next page of the prompt sequence is selected for another set Y (step 303). This selection for the next page may be performed automatically according to a script in the preferred embodiment, or may be performed in response to user input such as selection of a particular instance or other element.

  Once the next page is selected, the modified range for the attribute, ie attribute in set Y, runs the set decision engine in the attribute direction based on the remaining item set after the specified requirements for attribute set X are taken into account (Step 304).

  Next, the system will present the attributes in set Y with a modified range to the user along with an updated navigation window showing the number of items left after the request for set X is considered. The next page on which the page is currently displayed is presented (step 305).

  Next, an input indicating a measure of the attributes in set Y is accepted (step 306). An item / attribute preference score and request are calculated in response to this input (step 307). Once this score and request are calculated for set Y, the next page corresponding to attribute set Z is selected (step 308). A modification range for attributes in set Z is calculated based on the requirements specified for sets X and Y (step 309). The next page is presented with a modified range for the attributes in set Z and the navigation window is updated (step 310). Next, an input for the attribute set Z is accepted (step 311). Item / attribute preference scores and requirements for set Z are then calculated (step 312). An item score is calculated based on the attribute preference score and request for sets X, Y and Z and the above weighting rules (step 313). Finally, the resulting item arrangement ranked according to the item score is presented to the user (step 314).

  The process of FIG. 15 can be generalized to any number of attribute sets, such as the number set as described in the Summary section above.

  Thus, the present invention provides consumer personal decision support software that helps the average person to make complex and reluctant consumer product decisions or other complex decisions. This decision support software according to the present invention is a substantial and powerful type of expression class software tool known as a smart agent. However, the present invention can prompt the user to specify a personal profile for a particular product selection and intelligently apply that personal profile to provide decision support. Although consumers are confused by complex choices, there are also many areas of purchase where the present invention can assist in the decision making process. Buying new cars, investing in mutual funds, choosing the right university, buying real estate, deciding on family travel plans, deciding on health and medical choices, music and movies Selection and other fields are just a few examples of application areas of the scheme according to the invention.

  Overall, the present invention can significantly improve the efficiency of marketing and distribution of expensive consumer products and consumer products where complex decisions are made.

  The foregoing description of the preferred embodiment of the present invention has been given for purposes of illustration and description. It is not intended to limit the invention to the precise form disclosed. Obviously, many modifications and variations will be apparent to practitioners skilled in this art. The scope of the present invention is intended to be defined by the following claims and their equivalents.

FIG. 2 is a simplified diagram of the present invention in the context of an Internet-based decision engine server. It is a functional block diagram of the system of this invention. 1 is an architecture diagram of a decision engine according to the present invention. FIG. It is a figure of the prompt screen used with the figure user interface in this invention. FIG. 4 is a diagram of a checkbox prompt widget used in the graphical user interface of the present invention. FIG. 4 is a diagram of a numeric slider prompt widget used in the graphic user interface of the present invention. It is a figure of the radio button prompt widget used with the graphical user interface of this invention. It is a figure of the combo box prompt widget used by the graphical user interface of this invention. FIG. 4 is a diagram of a full notch slider prompt widget used in the graphical user interface of the present invention. FIG. 6 is an illustration of an “unwanted” half-notch slider prompt widget used in the graphical user interface of the present invention. FIG. 6 is a diagram of a “must have” half-notch slider prompt widget of the graphical user interface of the present invention. FIG. 4 is a diagram of a positive / negative preference slider prompt widget used in the graphical user interface of the present invention. FIG. 4 is a diagram of a positive preference slider prompt widget used in the graphical user interface of the present invention. FIG. 5 is a graphic user interface screen showing specific attributes in response to the exact set button in the window of FIG. 4 is a simplified flowchart of a user profile based determination process of the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Decision engine server 11 ... Display 12 ... Connection part 13 ... Internet 14 ... User station 15 ... User station 16 ... User station 17 ... Connection part 18 ... Connection part 19 ... Connection part 20 ... Display 21 ... Display 22 ... Display 25 ... Domain product set 26… Domain attribute set 27… User profile

Claims (24)

A method of generating a list of items from a product domain according to user preferences and requirements for an attribute A (n) of an item in the product domain for n equal to 1 to N comprising:
Defining a set of available choices for attribute A (n), where the set of available choices for attribute A (n) is within the product domain if n is equal to 1 Including available choices for all items of the item, and including available choices for items in the remaining product set if n is greater than 1,
Prompting the user to enter data indicating a preference or request for attribute A (n) in the attribute set according to the set of available choices for attribute A (n);
Storing the input data for the attribute A (n);
Determining, for i equal to (n, n−1,..., 1), a residual item set that satisfies the requirements for a subset of attribute A (i);
Repeating the steps of defining, prompting, storing and determining for attribute A (n) for n equal to 1 to N;
Assigning a score to an item in the remaining item set according to the input data indicative of a preference for an attribute. The method of claim 1 including ordering the items in the set of remaining items according to the score. The method of claim 1, comprising organizing the attributes into a sequence and providing the user with an indicator that indicates a position within the sequence during the prompting step. The method of claim 1, wherein the preference data includes a variable associated with a particular attribute specified by the user to have a degree of relevance for a selection within the product domain. The method of claim 1, wherein the request data includes a variable associated with a particular attribute specified by the user as to whether or not it is necessary for a selection in the product domain. The step of prompting for at least one attribute is positioned by a user at an input device on the bar to indicate a selection range represented by a bar displayed on the display and a selection displayed with the bar and within the range. The method of claim 1, including the step of displaying a graphic slider including a slide element to be operated. Organizing the attributes according to a hierarchical array; and assigning weights to the corresponding attributes in the attributes in response to the collected preference and request data and the position of the corresponding attributes in the hierarchical array The method according to claim 1. The method of claim 1 including generating attribute measurements in accordance with the input data. The attribute measurement for an attribute is
A numeric value for the numeric evaluation of the attribute,
A Boolean value for an attribute featuring a true / false indicator, and
An enumerated value for an attribute selected from the set or list;
Optional values for attributes included as standard features, which may or may not be included as optional features;
Said measurement and normalization routine having an individual measurement type selected from a set of measurement types including a ranged value for an attribute specified by a numerical range, and selected according to said measurement type 9. The method of claim 8, comprising calculating an item score for each item in the remaining item set in response to. The method of claim 8, comprising calculating an item score for each item in the set of remaining items in response to the attribute measurement. Assigning a score to an item, assigning a rule to the attribute specifying a response to missing input data for the attribute, identifying an attribute for which the input data is missing, and the identified The method of claim 8, comprising applying the rule to an attribute. Assigning a score to an item, assigning a rule to the attribute specifying a response to irrelevant input data for the attribute, identifying an attribute for which the input data is irrelevant, and the identified The method of claim 8, comprising applying the rule to an attribute. The step of prompting for at least one attribute is performed by a user at an input device on the graphic to indicate a selection range represented by the graphic displayed on the display and a selection displayed with the graphic and within the range. 9. The method of claim 8, comprising displaying a graphical tool that includes a positioned selector element. 14. The method of claim 13, comprising calculating a numerical value in response to the selected item. The method of claim 14, wherein the step of calculating a numerical value comprises applying a non-linear function to the position of the selector element on the graphic. Organizing the attributes into a sequence and providing the user with an indicator that indicates a position in the sequence and a count of items in the remaining item set during the prompting step. The method according to claim 1. A method for generating a list of items from a product domain according to user preferences and requirements for an attribute A (n) organized as a sequence of items in a product domain for n equal to 1 to N comprising:
Defining a set of available choices for attribute A (n), where the set of available choices for attribute A (n) is the product domain if n is equal to 1; Including available selections for all items in the list, and including available selections for items in the set of remaining items if n is greater than 1,
Prompting the user to enter preference data and / or request data for attribute A (n) according to the set of available choices for attribute A (n), wherein: A variable associated with a particular attribute specified by the user such that preference data has a degree of relevance for a selection within the product domain, and the request data is required for a selection within the product domain Including a variable associated with a particular attribute specified or not specified by the user;
Storing the input preference data and request data for the attribute A (n);
Determining a set of residual items for i equal to (n, n-1,..., 1) that satisfy the requirements with respect to a subset of attribute A (i);
Providing the user with an indicator indicating a position in the sequence and a count of items in the set of remaining items;
For n equal to 1 to N, with respect to attribute A (n), the above defining, prompting, storing, determining, and providing steps are repeated;
Assigning a score to an item in the set of remaining items according to the input data indicating a preference for an attribute, and ordering the items in the set of remaining items according to the score. Organizing the attribute set according to a hierarchical array; and assigning weights to the corresponding attribute in response to the collected preference and request data and the position of the corresponding attribute in the hierarchical array. The method according to claim 17, wherein: 18. The method of claim 17, comprising generating attribute measurements according to the input data and calculating the score in response to the attribute measurements. The attribute measurement for an attribute is
A numeric value for the numeric evaluation of the attribute,
A Boolean value for an attribute featuring a true / false indicator, and
An enumeration value for the attribute selected from the set;
Optional values for attributes included as standard features, which may or may not be included as optional features;
Having an individual measurement type selected from a set of measurement types including a ranged numerical value for an attribute specified by a numerical range, and said step of calculating said score comprises said step of said corresponding attribute The method of claim 19, comprising applying a normalization routine selected according to a measurement type. Said step of assigning a score comprises:
Assigning a missing data rule to the attribute that specifies a response to missing input data for the attribute, identifying the attribute that the input data is missing, and applying the missing data rule to the identified attribute;
Assigning to the attribute an irrelevant data rule specifying a response to irrelevant input data for the attribute, identifying an attribute for which the input data is irrelevant, and applying the irrelevant data rule to the identified attribute. The method according to claim 17, characterized in that The step of prompting the input of at least one attribute is performed by an input device on the graphic to indicate a selection range represented by the graphic displayed on the display and a selection displayed in the graphic and in the range. The method of claim 17, comprising displaying a graphical tool that includes a selector element positioned by a user. The method of claim 22, comprising calculating a numerical value in response to the selected item. The method of claim 22, wherein the step of calculating a numerical value comprises applying a fuzzy logic function to the position of the selector element on the graphic.

Images